Hi,
Used to? What has changed since, he wonders and ponders while searching Epping forrest for a fair amount of filaments so he'd burn this little villain for his satanic ideas...
Care to tell us before you reach Joan of Arc filamentary ( penny drops as I type) status?
Cheers and a lot of fun,
Used to? What has changed since, he wonders and ponders while searching Epping forrest for a fair amount of filaments so he'd burn this little villain for his satanic ideas...
Care to tell us before you reach Joan of Arc filamentary ( penny drops as I type) status?
Cheers and a lot of fun,
The switching frequency of switch mode PSUs is rising all the time. They now have to meet EMI regulations. It ought now to be possible to make one that is cleaner than your average (unregulated) linear supply yet for the same price. Subsequent (analogue) filtering would be cheaper/easier too.
Well, this just happens to be my area of expertise - smps design, that is - and I can't think of a better way to power a tube amp, with all of the different voltages it needs, most or all of which need to be isolated from the other. The noise? Fdegrove, your fears about noise and line pollution are well justified, but based upon rather dated technology. All the major players in SMPS control IC design have produced power factor correction circuits and the newer resonant mode controllers generate sine wave currents (or voltages, though that is much less common). Given that the minimum switching frequency for new designs these days seems to be 100kHz, it would be difficult to argue that any switching noise would contaminate the audio stream. The very real concern about loop stability of the error amplifier in a SMPS when powering solid state amplifiers (due to the very large amounts of negative feedback typically present) is not a concern in tube amps, and, in fact, having regulated and separated anode supplies eliminates motorboating without having to resort to smaller interstage coupling capacitors (the preferred technique back in the day, as I understand).
Much of the disdain for SMPS in tube equipment seems to be unfounded, knee jerk, or simply a prejudiced against all things silicon. The only good reason not to use one is that they are somewhat difficult to design and the voltage/current combinations most useful in tube amps are not going to be found in COTS* switchers!
* Commercial Off-The-Shelf
Much of the disdain for SMPS in tube equipment seems to be unfounded, knee jerk, or simply a prejudiced against all things silicon. The only good reason not to use one is that they are somewhat difficult to design and the voltage/current combinations most useful in tube amps are not going to be found in COTS* switchers!
* Commercial Off-The-Shelf
Rest assured I'm working on an all-discrete version.. slowly but surely.. think I need to grab another pack of transistors though, all the ones I have in my box don't seem to work anymore... (And no "", they seem to die on their own!)
It's a multivibrator (squarewave clock/PWM) > driver > output (TO-220 with some 400Vceo or more) driving an inductor, with high-speed diode to cap the flyback pulse, stepping 12V to whatever voltage.. current best result is 100V 30mA or so, at maybe 60 or 70% efficiency (mostly lost in the crappy drive circuits.. I need some PNPs
). 80% shouldn't be a problem...
Tim
(And no "", they seem to die on their own!)It's a multivibrator (squarewave clock/PWM) > driver > output (TO-220 with some 400Vceo or more) driving an inductor, with high-speed diode to cap the flyback pulse, stepping 12V to whatever voltage.. current best result is 100V 30mA or so, at maybe 60 or 70% efficiency (mostly lost in the crappy drive circuits.. I need some PNPs
). 80% shouldn't be a problem...Tim
jeffreyj said:
That's precisely why I think it will be hard to get one that's going to be quieter than my LCLC shunt regged linear supplies. I don't have the resources available any more to tackle such a design, and get good consistent performance out of it, that exceeds what I have now. For instance, the PCB design in these designs can be critical to performance, which for a DIYer is a huge PITA and expense. For a manufacturer, it's merely part of the development costs. Size and weight of my amps aren't an issue, and my linears have been refined to a very high degree.
So, as you have the expertise and tools at hand, how 'bout designing one for us? There's nothing off the shelf I've seen that will do. And I'll gladly test it for you. I'm not silicon-phobic, sometimes it's the best tool for the job.
Until there is actually a unit that physically exists, that is shown to be both technically and sonically superior and affordable, then a SMPS is not the best.
Thwack!*
*sound of gauntlet being thrown
EC8010 said:
Sorensen makes some very nice high voltage switching power supplies -- it can be done --
the problem is that the switching transients get multiplied in the power transformer (or at least they would in a perfect transformer) -- so you are back to using a slew controlled switcher chip with its somewhat lower efficiency (or some Rube Goldberg adaptation therof.)
i would venture to guess, however, that some of the Linear Technology PP gate drivers modified for off-the-line operation, driving a voltage doubler or tripler would compete quite favorably with a strictly linear supply.
A recent issue of audioexpress magazine featured a 300B amp with twin 5V switchmode supplies for the filaments, IIRC. So they are definitely creeping into use.
Maybe some commercial power supply company could design one especially for tube amp builders... maybe 350V at half an amp, and 6.3V at 10A... That would pretty much power most tube projects. It could be potted to look like a transformer!
I could definitley do with lighter amps, they are pretty darn heavy with all that iron!
Maybe some commercial power supply company could design one especially for tube amp builders... maybe 350V at half an amp, and 6.3V at 10A... That would pretty much power most tube projects. It could be potted to look like a transformer!
I could definitley do with lighter amps, they are pretty darn heavy with all that iron!
But can anyone reasonably expect a SMPS to contribute less switching noise than a pair of semi diodes? Coz lots of us don't dig the diode noise at all
I have once tried SMPS on the heaters of IDHTs with very disappoining results. What commercial valve equipment uses SMPSs? Apart from Nagra (reportedly not sounding too hot) is there anything else?
cheers
peter
I have once tried SMPS on the heaters of IDHTs with very disappoining results. What commercial valve equipment uses SMPSs? Apart from Nagra (reportedly not sounding too hot) is there anything else?
cheers
peter
analog_sa said:But can anyone reasonably expect a SMPS to contribute less switching noise than a pair of semi diodes? Coz lots of us don't dig the diode noise at all
Then use tubes. On said project of mine, I often use a 12AL5 or 17CT3 (convienient beacuse I have about that voltage (for heat) on breadboard and it's a mini 9-pin base) since I have no idea if any of my diodes can handle the frequency and voltage.
Anyway, to be more serious, there *IS* no switching noise, because it's at 100kHz. Err, uh, nevermind, audiophiles somehow mysteriously say they can hear that... riiiight...
Tim
My original thought was that maby it would be cheeper then a transformer. But the real point would be that it wighs less, and is more flexibel and could be suited for different constructions. (Easier that re-winding a transformer any way. )
Do any one know of any resources on the net that would help in the construction of a switched power supply?
/Johan Ch
)Do any one know of any resources on the net that would help in the construction of a switched power supply?
/Johan Ch
Bring on the Inquisition...
YES. Why not? We're all really proud of our carefully done LCLC smoothing, but the fact is that it's enormous because it has to operate at such a low frequency (100Hz or 120Hz). The inductors and capacitors needed to replicate that filtering at 100kHz are much smaller. Agreed, many SMPS produce noise, but that's simply because they're only made to be as quiet as is needed for their purpose. Despite the fact that computer supplies are cheap and cheerful, and that it's a noisy environment, it's still possible to put a soundcard in a computer. If that can be done, there's no reason why a SMPS designed for a valve HT supply shouldn't be a success. And as jeffreyj pointed out, they all have power factor correction circuitry, so they're actually a more benign load on the mains than a traditional "linear" supply.
Thinking of diode noise, can anyone think of a good reason (apart from efficiency) why a valve shouldn't be used as the rectifier in an HT SMPS? The output could then be followed with some LC smoothing to make it squeaky clean.
A dedicated HT SMPS isn't going to be something that can be made DIY, but how many of us wind our own mains transformers? We accept that, and buy them in, so why not a somewhat smaller black box with terminals?
analog_sa said:
YES. Why not? We're all really proud of our carefully done LCLC smoothing, but the fact is that it's enormous because it has to operate at such a low frequency (100Hz or 120Hz). The inductors and capacitors needed to replicate that filtering at 100kHz are much smaller. Agreed, many SMPS produce noise, but that's simply because they're only made to be as quiet as is needed for their purpose. Despite the fact that computer supplies are cheap and cheerful, and that it's a noisy environment, it's still possible to put a soundcard in a computer. If that can be done, there's no reason why a SMPS designed for a valve HT supply shouldn't be a success. And as jeffreyj pointed out, they all have power factor correction circuitry, so they're actually a more benign load on the mains than a traditional "linear" supply.
Thinking of diode noise, can anyone think of a good reason (apart from efficiency) why a valve shouldn't be used as the rectifier in an HT SMPS? The output could then be followed with some LC smoothing to make it squeaky clean.
A dedicated HT SMPS isn't going to be something that can be made DIY, but how many of us wind our own mains transformers? We accept that, and buy them in, so why not a somewhat smaller black box with terminals?
Re: Bring on the Inquisition...
That was my thought when I posted above.
A reasonable analogy, but I can get mains transformers wound in whatever I specify very reasonably, and there is a ton of surplus stock out there. A well made mains Tx won't buzz or do anything untoward for a century. Can that be guaranteed even for a decade for a switcher? And will we be able to get spares?
Switching at a high frequency also put out a lot of wideband garbage unless it's well taken care of by detailed design. And it will get into everything else in the system that isn't designed with that in mind.
I don't know where to get a suitable switcher and controller. If it can be made as good, it would be a boon to tweakers like me, so I can change voltages etc more easily and not have to call a neighbour so I can lift the amps back onto the stand. As this isn't for a walkman or a commercial product that must be shipped, weight and bulk aren't an issue for me at least.
So far a SMPS being better in audio is theory. SS and CD promised perfect sound in theory, yet they've failed to do so in practice. Why should a SMPS be any different? Poor experiences with the Tripath modules, RF sources from living near industrial environments and high power stepper motor controllers I worked with makes me skeptical. I'm willing to try, but with what?
EC8010 said:
That was my thought when I posted above.
A reasonable analogy, but I can get mains transformers wound in whatever I specify very reasonably, and there is a ton of surplus stock out there. A well made mains Tx won't buzz or do anything untoward for a century. Can that be guaranteed even for a decade for a switcher? And will we be able to get spares?
Switching at a high frequency also put out a lot of wideband garbage unless it's well taken care of by detailed design. And it will get into everything else in the system that isn't designed with that in mind.
I don't know where to get a suitable switcher and controller. If it can be made as good, it would be a boon to tweakers like me, so I can change voltages etc more easily and not have to call a neighbour so I can lift the amps back onto the stand. As this isn't for a walkman or a commercial product that must be shipped, weight and bulk aren't an issue for me at least.
So far a SMPS being better in audio is theory. SS and CD promised perfect sound in theory, yet they've failed to do so in practice. Why should a SMPS be any different? Poor experiences with the Tripath modules, RF sources from living near industrial environments and high power stepper motor controllers I worked with makes me skeptical. I'm willing to try, but with what?
Sch3mat1c said:
Again, please read Linear Technology's Application Note AN-70 for a discussion of how a low-noise switcher can be designed.
a 100 kHz switching frequency may save you some money on capacitors (higher switching freq requires fewer uF's for the same amount of ripple.) but the tradeoff is in the inductors -- the rapid transition from on to off means that peak currents are higher -- when the currents are higher the inductors heat, which can destroy the thing, or just waste energy in expended watts. So it's an optimization problem.
a poorly designed switcher will modulate the signal, a 100kHz signal will modulate the signal all the time. you have to put it in its own box, place ferrites on the leads in and out, chokes etc.
well, perhaps one of us should take a pair of like sounding D-76's and run one with a switcher, one with a transformer based supply.
wouldn't an OTL amp with a switching supply be a neat idea ?
Brett said:
The noise in linear supplies is at a very low level, but often has the troublesome 1/f distribution. Switching noise essentially starts at the fundamental of the switching frequency with a distribution of 1 over the odd harmonic (ie. - the fourier transform of a square wave). The residual noise (speckle, flicker, etc.) is often not even considered in SMPS because it is so far below the switching noise that it is - rightfully, some will say - disregarded. Honestly, I haven't given it much thought, but I will take some bandwidth-limited measurements of a few of the switchers I've got laying around here and see what turns up below 20kHz.
Well, it's funny you should ask...
I was trolling the "Tubes" section here because I'm about to design and build my first tube amp; a modest project, befitting my experience with such devices - a headphone amplifier. Probably a 6DJ8/E88CC cascode input driving a concertina splitter driving two push-pull White cathode followers, but I'm not quite sure on the topology just yet. However, I am very open to suggestions at this point (I have 2 6DJ8, 2 6SN7 and 2 12AU7A tubes to work with, hint, hint).
So, if you guys can come up with a good "functional spec," as is said on this side of the pond, I'd be happy to get to work on an SMPS design (seeing as I'll be needing to make one myself... heh). I'll need the following basic information:
1. The different (floating or ground referred must be specified!) output voltages and currents (eg. - 2 x 6.3V @ 2A each; 1 x 300V @ .1A; 1 x -100V @.01A).
2. Order in which they must be powered up! It is very possible to have the filaments idle at 63% voltage with the anode supply shut off, but the thought needs to be put into doing this in the beginning, or it quickly becomes a mess trying to add it on at the end!
I am going to put a firm cap on the total output power of 150W for now. Later on we can work on something for those monster mono-block tube amps I know some of you are dying to make
jackinnj said:
Well.... sorta, not really. The big problem with the capacitors is the ripple current rating: this is what dictates the capacitance value, not the standard dV/dt equation. The inductors are the least problematic design element in an SMPS - the transformer is the most. Finally, as switching frequency goes up, a greater proportion of the losses come from the switches, not the passives. Once the transition time exceeds 1% of the on time, no matter what the switching frequency is, the switches will waste too much power. However, making the swtiching speed too fast results in much higher leakage inductance spikes (for those topologies susceptible to such) and much higher loss in the gate drive circuit (several amps may be required to drive the gate/miller capacitance of a power FET at 100+kHz!!)
A poorly designed switcher will most likely do nothing but blow up!
The residual ripple of an SMPS will be inconsquential in an audio amp - after all, it will be well above human hearing. It is the spikes caused by the leakage inductance of the transformer which are most troublesome and can lead to sympathetic oscillation and/or ringing elsewhere in the amp.
Excellent ideas, both! It is easy to dismiss switchers on account of their inherent complexity and reputation for noise, but there are so many topologies to choose from that it is not even remotely correct to dismiss all switchers as "bad" for use in audio. The resonant-mode topologies, for example, produce extremely low amounts of noise at some penalty to the overall efficiency. Phase-control topologies allow recovery of the energy stored in the transformer's leakage inductance while also eliminating snubbers across the switches. Just because no switchers have yet been optimized for tube amp use does not mean that it can't be done (in fact, it is more likely that a knee-jerk "switchers are bad" mentality has prevailed, rather than anything else).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.